Thanks for reading! The best way to make sure that you don’t miss any of these conversations is to subscribe to The Good Fight on your favorite podcast app.
If you are already a paying subscriber to Persuasion or Yascha Mounk’s Substack, this will give you ad-free access to the full conversation, plus all full episodes and bonus episodes we have in the works! If you aren’t, you can set up the free, limited version of the feed—or, better still, support the podcast by becoming a subscriber today!
And if you are having a problem setting up the full podcast feed on a third-party app, please email our podcast team at leonora.barclay@persuasion.community
Francis Fukuyama is the Olivier Nomellini Senior Fellow at Stanford University. His latest book is Liberalism and Its Discontents. He is also the author of the “Frankly Fukuyama” column, carried forward from American Purpose, at Persuasion.
In this week’s conversation, Yascha Mounk and Francis Fukuyama discuss what the U.S. strikes in Iran mean for global and domestic politics, to what extent Israel is risking its relationship with the United States, and why we should be concerned about AI.
This transcript has been condensed and lightly edited for clarity.
Yascha Mounk: So I think we’re both struggling to follow all of the things that are going on in the world and to make sense of them. Let’s start with perhaps the most obvious topic: The supposed 12-days war in the Middle East, which Donald Trump declared over a few days ago. Is this actually the end? And what do you think it has accomplished?
Francis Fukuyama: Well, this is one of those questions that’s really impossible to answer at this point because it would depend on a lot of things that we don’t know—for example, exactly how much damage was done to this enrichment facility at Fordow. I think that in the short run, it looks like a pretty big success in many ways. First of all, the Israelis progressively managed to dismantle most of the forms of military leverage that Iran had, beginning with Hamas and then Hezbollah in Lebanon, then Syria fell, and then their initial strike really took out most of Iran’s air defenses, and so they’ve had pretty much air superiority for the past couple of weeks.
The one thing they couldn’t do is get at this deeply buried enrichment facility, but it looks like the Americans have taken care of that. So the Iranians are just in this extraordinarily weak position right now. I think the big uncertainty isn’t what’s going to happen in the short run, but rather what’s going to happen in the long run. Iran is a serious country, with 90 million people—even if you have regime change, it’s not clear that they’re going to give up on their nuclear ambitions.
One thing I think that this attack signals is that there’s a big distinction between nuclear and non-nuclear states in that you can do whatever you want to a non-nuclear state. It would be much harder if Iran actually had a nuclear program. I’ve all along felt that the reason they were going for this program is exactly that. Once you get nuclear weapons, it’s really hard for anybody to come and overturn your regime. If the Ukrainians had retained their nuclear weapons, I don’t think the Russians would have been able to do what they’re doing to them now. And so there’s going to be all these long-run consequences that aren’t going to be necessarily so pretty, meaning that a lot of countries are going to see this as a signal that they need to get serious about their own separate nuclear deterrence.
Mounk: So let’s talk about these two topics in turn. First, what is going on with Iran, and then the broader question of the danger from nuclear war and nuclear proliferation. When it comes to Iran itself, it does sort of seem as though the country is at the weakest point it has been internationally. I mean, it’s possible that the regime has some avenues for revenge, or to do damage either to Israel or to the United States, that haven’t yet become obvious, and that it’s somehow holding those in reserve, but that doesn’t appear likely at this point. It seems as though its ability to project power is really extremely limited at this stage. If Iran is weakened in those ways, how will that change the Middle East, and will that actually create opportunities for a more peaceful Middle East?
Fukuyama: Well, there’s another variable that you need to insert in there, which is the durability of the regime in Tehran itself. I think ever since the Mahsa Amini protests from 2022, it’s pretty clear that the legitimacy of that regime is really very low and it’s only its control over their power ministries, the IRGC and the militias that they control, that is keeping them in power. The regime itself is quite unpopular. And it would seem to me that this humiliation speaks to that. They’ve been trying to crack down because I think they themselves feel very vulnerable internally. So I don’t think you can answer your question about what this does for the larger Middle East until you answer the question of what’s going to happen internally in Iran.
Mounk: How do you see the different scenarios? Probably the most likely is that the regime is further internally weakened, but it is somehow able to hold on to power. And then I guess the other possibility is that the regime does fall in some kind of way, which presumably opens up a broad range of scenarios, including the most hopeful ones. Iran does have a more democratic history than some other countries in the Middle East, and after nearly 50 years of Islamist rule a lot of the population seems to be chafing at that. There are certainly strong secular voices in the country that would like to have greater rights for women, for example, and that see the way in which Iran’s prosperity has been damaged by the way in which the regime has turned the country into a sponsor of terrorism around the region.
But of course there may also be other extreme forces or voices within the country that are going to make a play for power as well. As always, when there’s a power vacuum, there may be a danger of civil war or other very dangerous outcomes. I guess, under the two different scenarios, what do you expect would happen over the next years and decades in Iran?
Fukuyama: You can’t ask that question. I have no idea. All of these things are possible. The country also has a lot of ethnic divisions and you could have breakaway parts. There’s just no way of answering that question at this point. We can hope for a good transition, but our experience with regime change in that part of the world is not all that happy. So I just think it’s useless to speculate on that stuff.
Mounk: That’s why I ask you the questions and I hope that you answer them, but you’re right that I certainly don’t feel capable of answering them.
How do you think about this under conditions of uncertainty? I think it’s easy to feel torn when you have a horrible regime—and the Islamic Republic of Iran certainly is a horrible regime, in terms of what it’s done around the region, around the world, and in terms of how it’s treated its own citizens—to feel torn between the hope that this regime falls, that it gives way to something much better, and the apprehension that it could give way to something worse, or it could give way to the kind of chaos that leads to a lot of death.
Now, there’s a kind of political lesson: trying to lead wars of regime change that are designed to change the political structure of another country is probably a foolish idea just because of how unpredictable historical events are under most circumstances. But another is sort of just, I guess as an armchair observer of history, I mean, where should our sympathies be? Should we hope that tomorrow a grand revolution breaks out and the regime falls? Or should we fear that? Or should we do both at the same time? Or should we think like Immanuel Kant supposedly did about the French Revolution—that on the one hand, from a logical or moral point of view, we don’t want it to happen because we think that what could result is bad. And on the other hand, we can’t contain our joy that something better may be around the corner. What do you think about those things?
Fukuyama: I think that it would be a great thing if this regime actually fell. I think some of the alternatives are very problematic, especially for people in Iran. But it’s a little bit hard to think of regimes that would be terribly much worse than this one. I guess the one thing you have to say about Iran that really makes it very different from Iraq is that it’s a real country. And this is something I think people didn’t appreciate in the earlier stages of this conflict. Statehood in Iran goes back millennia. It created one of the first states anywhere in the world and that state was consolidated centuries before states were consolidated in Europe.
A lot of Iranian friends of mine keep pointing to the fact that we picked the wrong side in the Iraq-Iran war. Iraq had never been a country, really. Part of the chaos that ensued after the American invasion was the fact that it wasn’t a country. Whereas Iranian statehood is really not contested. So the question is who manages to come to power. It could be the military. I mean, it could be the IRGC. Well, I mean, they’ve been the basis of the regime’s power up till now, and it could be that they’ll simply take over. I’ve heard Iranians say that if a more democratic Iran comes to power, it might become one of the most secular countries in the whole world because they’ve been living under this theocratic dictatorship ever since 1979 and most Iranians really hate it. The range of possibilities is very great, but I don’t think anyone’s going to shed any tears if this particular regime falls.
Mounk: I’ve seen a lot of people compare the current conflict to the Iraq War for some obvious reasons, both that Iran and Iraq are obviously adjacent countries, and that there’s been some rhetoric around regime change, both from Israel and in some ways from Donald Trump. But it seems to me that the disanalogies are at least as important. One disanalogy, as you pointed out, is in the nature of those two states: that Iran has a much longer history as a state, and that even though Iran is quite ethnically and to some extent religiously diverse as well, I think it has a much clearer ethno-religious majority than Iraq ever did, which makes the kind of civil war that Iraq ended up in unlikely in Iran.
The other perhaps is that it seems to me that the goals that both Israel and the United States are pursuing in Iran this time around are quite different. At the time of the Iraq war, the United States obviously planned from the beginning to have boots on the ground for some amount of time. They didn’t anticipate it being as long as American troops stayed in Iraq, but they certainly thought that they would physically invade the country. That was part of the plan from the beginning. And there was this kind of idea that America would impose an American-style or Western-style democracy. That would be a shining example to other countries in the Middle East. And it was kind of buttressed by a certain hopeful set of ideas about how easily you can export democracy to other places.
It seems to me that even though Netanyahu and Trump may in various ways hope that the humiliation of Iran might be the end of the regime, they’re certainly not proposing to send ground troops to occupy Iran, which would be a vastly more complicated, costly and deadly undertaking. I don’t think that either of these countries really think that they’re going to go and send over advisors and write the new constitution of a country. So, is Iraq actually a useful metaphor in terms of how to think about this, or is the nature of this conflict very different?
And that of course also goes to some domestic questions we might ask about how this will impact the popularity of Donald Trump. It’s easy to forget that the last Republican president before Trump was George W. Bush, and Bush’s presidency went south in good part because of the failure in Iraq. Trump claims to have learned that lesson—he said in his inaugural address that he was going to be measured in part by the wars that he doesn’t start. I think there’s a part of Trump’s base that was very worried that he was giving up on those lessons, as he was clearly contemplating joining this war against Iran. I think perhaps some people on the left hoped that in a way this might deal a blow to Trump’s popularity.
But it seems to me that he’s taken quite a different approach to this, which is to send in a bunch of big bombs, do some actual damage to the Iranian nuclear program—and it’s a victory very very quickly. That’s a very different kind of war and it might actually allow Trump to go around saying, I’m the first American president in twenty-five years who’s won a war, which could be quite popular with the American public, actually. So I’m interested both in substantive terms of what you think about this comparison to Iraq and, in more American domestic politics terms, how do you think this will impact Trump’s presidency?
Fukuyama: Well, I think that actually we did learn a lesson from Iraq that is playing out right now, which is don’t get involved in the domestic politics of countries in the Middle East and think that you can shape their outcome. I think that’s something that we’ve learned very painfully. And Trump is certainly not going to try to do that. Israel, even less so, has the ability to really shape events in a successor regime. And that’s a healthy lesson.
I don’t know of anybody who thinks that we’re gonna repeat our efforts to actually create a democratic regime in Iran the way we tried to do in Iraq. That’s just kind of lunacy at this point. The impact on American politics is one of those things that puts me in a bind, because it does seem to me that Trump could come out of this looking very, very good. There’s a kind of minimal willingness to take risks, that he says he’s going to do something, he actually does it, it produces a result that actually looks pretty good in the short run, and I think it’s going to increase his popularity and legitimacy immensely. I think that’s not good for the United States, but it’s a good outcome in terms of foreign policy.
It’s funny because I kind of thought that one of Trump’s big vulnerabilities is that he’s not a strategic thinker and he’s never been tested in a genuine crisis and he was not likely to come out of it looking good. He could come out of this one looking pretty successful. I think that in the longer run, if there’s not a kind of orderly succession that leads to stability in Iran, if it continues to be a source of terrorism and draws other countries in and that sort of thing, he could look foolish. But I just think that in the short run, and in a cycle that is attuned with American politics, he could actually benefit from this quite substantially.
Mounk: As we’re recording this It looks as though the ceasefire may break down, though it may also very well get back on track. It’s a little bit hard to know and of course one question is, what will the situation be in two days? Another question entirely is, what’s going to be the case in two weeks or two months? But if Trump is able to say to the American people, I took decisive action, I eliminated a nuclear threat both to us and to our ally, and then I got out, and that’s it, then that’s obviously going to strengthen his domestic standing. I guess the other question to ask is about Israel, and there too there’s a question about whether the apparent success of his mission will strengthen Benjamin Netanyahu, who was politically on the ropes once again before he decided to start this war with Iran. He now may very well be strengthened in his domestic politics. And there’s a broader question. Israel, on the one hand, now looks stronger and safer than it has in a very long time. It has brought under control not just Iranian proxies in Lebanon, obviously in the Gaza Strip, but also very significantly weakened an Iranian regime that was extremely hostile to the country.
At the same time, it has lost a lot of support in the international community, including among its allies. You see Emmanuel Macron, the president of France, suggesting that he will soon recognize a Palestinian state, which a number of European countries like Spain have already done in recent years. And it just looks increasingly as though Israel is staking its midterm and long-term survival on the ability to militarily dwarf its neighbors for the indefinite future. Even though it clearly is capable of doing that right now, I think there’s just real questions about how wise a bet that is strategically in the long run, especially if the actions it takes, thanks to its military superiority at this point, are part of what erodes the alliances and partnerships that have kept it safe in the past.
Fukuyama: I think that in the case of both the United States and Israel, the big danger right now is to be, as Stalin said, dizzy with success, having succeeded in defeating enemies unexpectedly. I’m sure Netanyahu and the Israelis feel pretty unconstrained that they can use their military power without any real fear of blowback, certainly in the short run. Trump may feel that way also and may try to use that American power in other theaters in ways that aren’t gonna be as successful. So I think that’s kind of the main problem.
I think for Israel, they never had that strong support in Europe in the first place, so the real issue is that the attitudes of young Americans towards Israel are really, really different and they’ve been really damaged by the war in Gaza. At the moment that’s not going to endanger the kind of military and economic support that the United States gives Israel, but in the long run, it’s going to erode that support. It’s already the case, I think, that Israel has been sucked into the partisan divide in the United States with very strong support coming from conservative Republicans and a lot of active hostility coming from the left wing of the Democratic Party. And I don’t see that reversing at any time in the future. I just think Israel is going to have to live with that, and therefore they’re going to fear the election of a Democratic president that may go back to a much more neutral and much less supportive posture towards them. So I would say that’s the main danger that they face at the moment.
Mounk: Yeah, and it’s one thing for Israel to lose the support of various European countries that have always been rather lukewarm in their relationship to Israel. But if in the future you had a U.S. president who really is very critical of America’s alliance with Israel and is actually willing to cut off some key elements of it, that would pose a much more significant danger to the country. And I agree with you that some of the actions that Israel is taking today are because its leadership thinks that’s what it takes to keep the country safe in the short run.
Fukuyama: One long-term thing that the Israelis need to worry about, among the things they could do if they really get dizzy with success, is something that I think a lot of right-wing Israelis have wanted for some time, which is basically to declare sovereignty over the West Bank. And at that point, they’ve got two choices. They can actually try to expel the Palestinian population. They already want to try to do that in Gaza. Or they could just, you know, frankly admit that Palestinians are there but they’re not going to have the rights that Arab Israelis have enjoyed up to this point. And at that point, Israel actually begins to look like what its enemies have accused it of being, which is an apartheid state where you basically deny the premise of equal dignity of all people, regardless of ethnicity, and you say, well, the Arabs are at a lower status.
I think that in the long run, this is very bad for U.S.-Israeli relations because essentially if you’re not Jewish, why do you support Israel? For more secular Americans, I think a lot of that had to do with the fact that Israel was the only genuine democracy in the entire Middle East, and that as a fellow democracy, we owed Israel support.
But if they move to being, more formally, a very illiberal democracy where the democracy only applies to the Jewish part of the population, it raises a real legitimacy question for them. And I think that it’s—I certainly wouldn’t want to go there if I were an Israeli—but that certainly is the impetus for many of the people on the Israeli right.
Mounk: Let’s assess a little bit where we’re at in the United States. We’re about five months into Donald Trump’s first term, so there’s a lot of it yet left to go, about three years and seven months. So we’re still very much in the early stages. How worried are you now compared to the last conversations we had about Trump’s presidency? I think on the one side, you see some of the scenarios that people had warned about starting to play out. One of the concerns that a democracy watch has had from the beginning was that there might be some form of clash between the forces of order and protesters, and that that might lead Donald Trump to call out the National Guard in order to “restore order,” and that that could get out of hand very quickly.
We’ve seen a—so far comparatively mild—version compared to the fears of that play out in Los Angeles, but it very clearly is a sign that those exact scenarios may be in the offing. On the other hand, it feels a little bit as though the shine has come off the Trump presidency. He actually continues to be relatively popular. When I checked just before recording this, in Nate Silver’s average, he was down by about five or six points in the approval ratings. So he’s more unpopular than he is popular. But those are not terrible numbers. Those are better numbers than a number of presidents had at the same stage in their presidencies. But certainly the sense which the first hundred days of the presidency gave, this feeling that there is this master plan for how to broaden the power of the presidency for all of the things that the executive is going to do, for how it’s just going to completely transform American life and American politics—that feels like it has run aground a little bit.
It’s perhaps inevitable that the kind of speed and the radicalism of the first hundred days of the administration can’t be kept up for four years, but it certainly has felt in the last 50 or so days as though the speed with which things are happening is slowing down, as though some of roadblocks from the federal courts, which have held up so far, are making it harder for the administration to actually go full speed ahead with some of the that they’ve wanted to do. So I’ve been a little bit confused about how to think about the balance between those two different things. On the one hand, I think the growing indications are that Trump really is very impatient with some of the limits on his power. On the other hand, the fact that the speed, and frankly the advocacy, with which the administration moved in the very first days of its reign seems to have slightly subsided. How are you making sense of this moment, Frank?
Fukuyama: I think it’s subsided. I think, however, I’m assuming that his popularity is going to recover pretty substantially as a result of the attack on Iran. That might fade away over time, but I do think that he’s helped himself, in that just seeming decisive as a president that can get things done. It’s still going to depend on a lot of the factors that we talked about earlier, particularly the economic ones. The 90-day tariff deadline is coming up in a couple of weeks, but as far as I can tell he hasn’t really negotiated any bilateral deals with anybody. So then he faces this prospect of, does he chicken out or does he double down? If he doubles down, you’re going to have a lot of negative economic consequences. You’re also seeing him back down on a lot of the deportations, because it’s really been devastating, like in meat packing or hospitality or a lot of other domains in the economy, and he’s kind of recognized that that’s hurting people who are in his base and therefore he’s been granting exceptions and cutouts and so forth. So it could be that he will actually chicken out on a lot of the harsher aspects of his threatened policies.
Mounk: It’s one of those strange things where it seems like the less he does, the more popular he might end up being. The stock markets are now up since the beginning of the year. They’ve been up since Trump took office. They’re not quite at record highs, but they’re doing very well. The American economy may have slightly suffered because of the unpredictability and the chaos instilled by Liberation Day, but it seems to be chugging along at a somewhat decent pace. So if he somehow declares victory after a hundred-day pause and passes off the five and a half bilateral trade deals that he’s instituted and says, this is all I was hoping for and there we go, then he might be able to benefit from a relatively thriving economy and come out of this relatively well. If of course a few weeks from now he says, well, we have these five and a half trade deals and congratulations to these countries, but everybody else is back to having these extremely high tariffs, and the stock market plunges again and all of the Wall Street people who seem to be betting on the idea of TACO, of Trump always chickens out, look foolish, then that ironically might then harm Trump quite a lot.
Is it too simple to say that if Trump sits back and doesn’t do very much for the next few years, he might actually end up being quite popular and then ironically be able to expand his power in some of the ways that he wants, whereas if he pursues some of his ideological goals, he might end up being unpopular and that might actually make it much harder for him to reach those other goals? There’s a kind of tension here between his ideological views and his sort of desire to expand his power.
Fukuyama: I think that’s right. If he does chicken out, that’s probably good for him. It’s going to take away this major threat of tanking the economy. And like in the first term, people will say, well, he threatened a lot, but in the end, it wasn’t so terrible. And that might ease the pressure on some of the institutional changes that he’s seeking in terms of the power of the presidency. So yeah, I think that that’s definitely a threat.
Mounk: How about some of the immigration things? There’s obviously significant concern that the Trump administration was ignoring some lower court rulings on immigration. They seem to mostly have desisted from that strategy. The person that they deported to El Salvador under supposed suspicion that he was a member of a gang is back in the United States—there’s an ordinary set of court proceedings about him. Mahmoud Khalil, the Columbia University student who was being held, has been released and is in New York. The federal government is trying to get him deported, but they’re doing that in an ordinary court process while he is with his family in New York City. So it seems as though on the one hand the administration is trying to take very tough action on immigration, but on the other hand seems, for now at least, to have decided that they’re not at this moment seeking a full frontal confrontation with the courts where they’re saying we’re not going to listen to what the courts tell us or to ways they’re trying to constrain us on immigration. How do you think this immigration policy is playing out, and how will that shape Trump’s presidency?
Fukuyama: The problem with comprehensive immigration reform is that at all the moments it’s been tried—the last big one was during the George W. Bush administration—there was the question of so-called amnesty or giving undocumented people a path towards citizenship, which was an absolute no-no for conservative Republicans. I think the experience of what’s going on in several of these industries that are highly dependent on immigrant, undocumented immigrant workers is actually beginning to make a case that conservatives may recognize that you’ve got these people, they’re not criminals, they’ve been working hard, they’ve got families, and we’re really dependent on their labor—that might set up the ground for moving in the direction of some way of actually legalizing them.
Now, I think if Trump were really smart and strategic about this, he would take that up. And you might actually get this miracle of something like comprehensive immigration reform. The trouble is that there are other people, like Steve Miller and Steve Bannon and other people in his orbit, who just seem hostile to foreigners. They’ve been doing this stuff that to me is just inconceivable, like telling green card holders not to leave the country because they may not be allowed back in, by saying that they want to deport even people that are legally in the country, or making their life as difficult as possible. Or threats against universities not to accept foreign students. Where does that come from except for this just very nasty nativist desire not to have foreigners in the United States? And so there’s also that side of this administration and I think they may well be at war with one another. And so again, it’s a little hard to know which of these forces is going to be dominant.
Mounk: One of the interesting things about the first five months of the Trump administration in my mind is the way in which he’s really failed to deliver on what I call aspirational populism. You know, when you look at how Trump got reelected in November of 2024, this was obviously in good part by expanding the traditional base of the Republican Party to include a lot of young voters, a lot of Latinos, a lot of other demographic groups that historically have not voted for the Republican Party in large numbers. And I’ve argued, for example, in conversation with Arlie Hochschild, that one way to understand that is to tell a different kind of deep story of what motivates those voters.
In 2016, we used to go back and say, these are the old white men who are just nostalgic for the past and want things to go back to how they were. I was always somewhat critical of that reading of what was happening, but it really is untenable when you look at who voted for Trump in 2024. I think a lot of those were people who were saying, hey, my life is better than that of my parents and better than that of my grandparents. And I want somebody who gives me even more opportunity. I want somebody who allows me to start a small business and really thrive and save a bunch of money on taxes if I’m doing well. I want somebody who perhaps will take America to Mars with this alliance with Elon Musk. It’s just striking to me that Trump has mostly given up on those voters. He’s given up on those voters with a budget that is very unpopular because it hands out relief to the very richest Americans, but is basically negative on the incomes of the vast majority of Americans.
Fukuyama: Including a lot of the rural poor white voters that are really the core of the MAGA base.
Mounk: That’s an interesting point, actually. So it doesn’t do anything for those young Latinos, for example, who moved into Trump’s column in November 2024. But nor does it actually support the kind of core constituency that was with Trump from the beginning. That's a very good point. Obviously, his alliance with Elon Musk has broken up in an interesting way, this idea that he’s going to have the best and brightest from Silicon Valley run this aspirational playbook about what the American government could be and what the United States could achieve—that seems to have mostly gone away. And then on immigration, as you’re saying, there’s a majority of Americans who want better control over the southern border, who might be open to more energetic deportations, and that includes a lot of Latino voters. But the apparent hostility to most foreign-born Americans that you feel emanating from people like Stephen Miller and the administration, it’s just a minority position in this country. And it certainly is a minority position among some of those new voter groups that had moved towards Trump. And so it is interesting that what I think a populist movement in America in 2025 in some sense “wants to be,” which is to appeal to this multiracial working class constituency, is not something that Trump has been very effective at actually putting into political practice in the first part of his presidency.
Fukuyama: No, that’s right. I guess the question is whether his basic political instincts are going to lead him to a more moderate position, recognizing that he does need to hold on to those new voters. And that’s one respect in which he is not an ideologue. I mean, he’s just kind of infinitely opportunistic. And he may perceive that this is not a winning position. But we’ll have to see.
Mounk: I also wanted to ask you about a completely different topic. You've written very interestingly both about how much you have enjoyed using AI, how powerful a tool you think it is, and, more recently, about how you've come to be much more concerned than you had been previously about the danger that AI poses, about the ways in which it might really elude the control of humans, in particular because of something called agentic AI. Tell us about why you've become so much more concerned of late.
Fukuyama: I constantly get asked by journalists my opinions about AI and I have been avoiding talking about that for a very long time, but I’ve started to actually read and think much more seriously about the problem. And I guess that I now see why the more extreme fears that people have are actually something we ought to worry about. And in fact, we’re very far from, I think, actually comprehending what the nature of that threat is. There are two sorts of dangers posed by AI.
One is simply an acceleration of what we’ve seen with other forms of technology, which is that AI will allow targeted advertising, it will allow cyber crime, it will allow all sorts of deep fakes that will further undermine the trust that people have in the media, in normal democratic discourse, which really requires agreement on basic facts. All of that is happening. I’m really struck by the fact that, in my Instagram feed, I would say that 50% of the videos there are deep fakes. Maybe it just reflects the kind of stuff that I tend to watch, because the algorithm gives you more of whatever you start watching. But for example, there’s this genre of videos where some guy in a souped-up pickup truck runs over a bunch of police officers. And first of all, it’s kind of creepy that there’s a segment of the population out there that really wants to be able to run over police officers in these souped-up trucks. But they’re obviously fake videos—they look very realistic at first, but they’re obviously fake. There’s another genre of cruise ships crashing into one another and ripping big holes in their sides of each other. There’s another genre of tourists on a beach and then a tsunami comes and washes them all away. So that’s already happening. And that is going to accelerate.
But the real problem is these fears that are more existential—that AI could wipe out the human race at a certain point. This is the one I never really understood because it seemed to me that you’d always be able to hit a kill switch for this type of AI. But it does seem to me now that I see much more clearly a path by which you could get to that point. And it could be coming much, much more quickly than people recognize.
Mounk: That’s to do with this idea of agentic AI, right? How is agentic AI different from just me typing a question into ChatGPT?
Fukuyama: So, agentic AI is where you actually delegate authority to an AI to make decisions that would normally be made by a human being. And in a certain way, we already do that, right? We have a lot of systems that monitor traffic and if they see something outlandish appearing on social media, without any human intervention, they correct that.
But the issue comes out of my thinking about the problem of delegation. I think that the problem of delegation is the central problem in politics. Every hierarchical organization needs to delegate authority because the problem is that it’s the agents in a principal-agent relationship that really oftentimes have the greatest knowledge and the ability to act quickly and so forth. If you try to control everything from the top, you’re simply not going to have a successful organization. And this is true in virtually every human organization. So we are increasingly delegating authority to agentic AIs to make decisions on our behalf.
I just think that that’s going to grow because they know much more. If you hit AGI, artificial general intelligence, which some very smart people think is going to happen within the next few years, you’ll basically be having these agentic AI agents that are as capable and much, much more knowledgeable than any human being. But the other kind of creepy thing that I don’t think people have quite taken on board is that, at the cutting edge of AI today, the most sophisticated AI systems, the ones that people are investing tens or hundreds of billions of dollars in, are not being programmed by anybody. It’s not like you got a bunch of software engineers that are saying, okay, we want a machine that will do X, Y, and Z and not do A, B, and C. The metaphor that’s used is that they’re growing, meaning that these machines are basically programming themselves, that they’re given a base set of instructions and they have a lot of knowledge that they can take in. They take in more knowledge, they try things, and if it works, they do more of that, and if it doesn’t work, they stop doing it. But that’s not being controlled by anybody, it’s controlled by those machines themselves.
So when you talk about delegating to an AI, you don’t actually know whether you can trust the AI. And there’s all these scenarios. For example, if AI start talking to one another, they invent a language that only they understand. So you’re not actually going to have human beings that know what they’re saying to one another. There are scenarios where they can exfiltrate themselves so that someone decides that this AI is doing something dangerous and you want to shut it off. But they’ve actually already inserted themselves into other computer systems that you don’t know about. And I think that there are these anecdotal cases where these AIs are doing stuff that nobody expects. I mean, this is the experience I think that we’ve all had where ChatGPT comes up with stuff that is very surprising. You don’t know where they got this idea, but this is something that is happening to an increasing extent with these machines.
It just seems to me that if you know anything about the way that human organizations work, you know that over time they delegate more and more authority to lower levels of hierarchy. And this is already happening with AI and it’s going to continue to happen. You’re already seeing this in terms of military affairs. So we’ve said for some time now that there’s always going to be a human in the loop before a machine can kill another human being. That’s being violated as we speak. The drones that are being used by Russia and Ukraine are controlled by AI programs where there is no human in the loop—they can recognize a hostile target and they can just decide on their own to destroy it. So you’re already seeing examples of delegation of the use of lethal force, as we speak. They refer to this as the loss of control problem with advanced AI. And I just don’t see how this isn’t going to get a lot worse as time goes on.
Mounk: I’m really struck in this moment by how AI is both kind of in the background of every conversation and still, I think, actually underestimated as a force that is just going to completely reshape human life and society. And I’m struck by how many people have this reaction of, I asked ChatGPT to look up some quote and it hallucinated an answer, so it really can’t do anything—the number of people who really are sleeping on the extent of the capacities that these systems now have. I think there’s some real cope of people saying, these systems aren’t really intelligent because they’re just these stochastic parrots. They’re just these algorithms. There’s just this kind of physical base for how they give these answers and so therefore it’s not real intelligence, which of course ignores the fact that at some level of description, which we are still not fully capable of rendering about our own brains, those are machines made of matter as well. Some of those arguments that are used to draw the conclusion that supposedly AI is not really intelligent could actually be applied just as well to humans, and I think would be just as misleading in the case of humans.
I think this dimension that you’re drawing our attention to of agentic AI is something that is just outside of the experience of most users. That is not how most people interact with ChatGPT or whatever other interface they may have experimented with. And so it’s really outside of their imagination.
Fukuyama: One thing I didn’t mention is that it’s going to be really hard to do anything about it because of geopolitics and also just competition within that whole sector that the advantages of getting to AGI first are going to be really, really huge. Both China and the United States are racing towards that. And the constraints that we want to put on the degree to which we delegate authority to AI agents is only going to slow us down in that race. That’s why I’m now kind of pessimistic about whether we can control this, because the countervailing pressures to get there first are already very, very powerful. So I don’t know what’s going to happen, but I think that this is going to be the chief worry in the coming years—that these systems are so powerful and that we are actually losing control of them.
Mounk: If people want to understand this better, if listeners are a little bit skeptical of what you’re saying or they just feel lost when it comes to AI, what can they do to experiment with AI by themselves to give them a sense of those capacities? What can they read? What can they listen to? What is a good way to get smarter on this topic?
Fukuyama: If you go on YouTube and you search for a couple of interviews with Geoffrey Hinton. He’s called one of the godfathers of AI because some years ago he designed some of the fundamental building blocks of what’s become things like ChatGPT and he’s got an enormous amount of credibility. He lays out very clearly what some of the dangers are, including these more existential threats. And I think that’s probably a good place to start because you go to someone like Sam Altman, it’s not helpful because he’s got so many mixed motives. He’s got a commercial interest. He wants to downplay the dangers, and I don’t think that’s going to be very reliable. Someone like Hinton, his career is largely done. And I think that he is much more credible in terms of talking sensibly about what the future may hold. So that’s my recommendation.
If you have not yet signed up for our podcast, please do so now by following this link on your phone.
Email: goodfightpod@gmail.com
Podcast production by Mickey Freeland and Leonora Barclay.
Connect with us!
X: @Yascha_Mounk & @JoinPersuasion
YouTube: Yascha Mounk, Persuasion
LinkedIn: Persuasion Community
Share this post